Code to run analysis and generate figures for McNeall et al. (2023) “Constraining the carbon cycle in JULES-ES-1.0”

## Warning: replacing previous import 'lifecycle::last_warnings' by
## 'rlang::last_warnings' when loading 'tibble'
## Warning: replacing previous import 'lifecycle::last_warnings' by
## 'rlang::last_warnings' when loading 'pillar'

Failure analysis

Failure analysis marginal plots

Are there hard thresholds after which the model always fails? (Question from review comments). Run failures are in t

Visualising the ensemble range

It’s important to remember that the design of the experiment is multiplication factors of the original parameters. This might be important for the “hold” value in a sensitivity analysis, as the “standard” value and the median value of the ensemble will not be the same.

Wave00/Wave01 Ensemble behaviour in key (constraining) outputs.

Global mean for the 20 years at the end of the 20th Century. There is still a significant low bias on cVeg output.

What proportion of wave01 fall within Andy Wiltshire’s constraints?

Just under a third. Points at a significant model discrepency in cVeg

Of the 400 members of the wave01 ensemble, 128 pass Andy Wiltshire’s Level 2 constraints.

## [1] 128
## [1] 0.32

Exploring summary of constraint: Cumulative NBP

Carbon budget data

Section 2.5 in Friedlingstein et al. describes how the land carbon sink is estimated. IPCC AR6 Chapter 5 states: “The net land carbon sink is taken as net biome productivity (NBP) and so includes any modelled net land-use change emissions. Further details on data sources and processing are available in the chapter data table (Table 5.SM.6).”

The GCP says: “The land sink is the average of several dynamic global vegetation models that reproduce the observed mean total land sink of the 1990s.”

AR6 Chapter 5 states: "The land carbon cycle components of historical ESM simulations show a larger range, with simulated cumulative land carbon uptake (1850–2014) spanning the range from –47 to +21 GtC, compared to the GCP estimate of –12 ± 50 GtC (Figure 5.23b).

How much are all select output constrained?

Constraint of output at level 2

Proportion of the initial range of model output that is covered by level 2 constrained ensemble (%):

##                   npp                   nbp                 cSoil 
##                  26.6                  55.1                  48.3 
##                  cVeg          lai_lnd_mean            rh_lnd_sum 
##                  23.5                  60.9                  26.0 
##          fLuc_lnd_sum      fHarvest_lnd_sum     treeFrac_lnd_mean 
##                  51.5                  60.7                  88.3 
##    shrubFrac_lnd_mean baresoilFrac_lnd_mean    c3PftFrac_lnd_mean 
##                  85.4                  52.9                  58.0 
##    c4PftFrac_lnd_mean 
##                  43.9

Pairs plot of 2d projections of constrained output space.

Constraining to level 2 with the emulator

## [1] 12.334